Liver Fat Quantification In Medium and Large Phantoms: Photon-Counting CT Vs Dual-Energy CT Virtual Monoenergetic Imaging
Abstract
Purpose
To evaluate the relationship between fat volume fraction (FVF) and Hounsfield units (HU) in unenhanced fatty lesions and to identify virtual monochromatic imaging (VMI) settings that minimize FVF quantification errors, by comparing photon-counting CT (PCCT) with dual-energy CT (DECT) across different patient sizes.
Methods
Six fatty lesions (FVF 5%–40%) were embedded in an anthropomorphic liver within a medium abdomen–pelvis phantom (25×32.5 cm²) and a large phantom (31×39 cm²) created by adding a soft-tissue–equivalent layer (CIRS). Phantoms were scanned on Siemens NAEOTOM Alpha using an abdomen–pelvis protocol (120/140 kV; CTDIvol 8.51 mGy [medium] and 12.8 mGy [large]; QR40f; QIR 3; slice thickness 0.8 mm). Spectral images were analyzed across VMI energies (40–190 keV). DECT scans were performed on a GE Revolution system using fast kV switching (80/140 kV; 14 mGy [medium]; 21 mGy [large phantom]), generating VMI images from 40–140 keV. HU values from three repeated scans were measured at each VMI energy. Linear regression between HU and known FVF was performed, and quantification accuracy was assessed using root-mean-square (RMS) error of predicted versus true FVF.
Results
For medium phantom, minimum RMS errors were 4.8% at 100 keV on DECT, and 1.9% (120 kV, 56 keV) and 1.8% (140 kV, 68 keV) on PCCT. For large phantom, minimum RMS errors were 15.3% at 130-140 keV on DECT, and 4.6% (120 kV, 105 keV) and 3.2% (140 kV, 75 keV) on PCCT. PCCT demonstrated improved FVF–HU linearity, particularly in the large phantom.
Conclusion
The FVF–HU relationship depends on CT technology, beam energy, VMI energy, and patient size, with distinct optimal VMI energies for accurate FVF assessment. Compared with DECT, PCCT substantially improves FVF quantification, achieving RMS errors of <2% for medium and <5% for large phantoms at 39% lower radiation dose.